CN113935443B - Heterogeneous equipment multi-domain joint fault prediction method, device, terminal and storage medium - Google Patents
Heterogeneous equipment multi-domain joint fault prediction method, device, terminal and storage medium Download PDFInfo
- Publication number
- CN113935443B CN113935443B CN202111558674.XA CN202111558674A CN113935443B CN 113935443 B CN113935443 B CN 113935443B CN 202111558674 A CN202111558674 A CN 202111558674A CN 113935443 B CN113935443 B CN 113935443B
- Authority
- CN
- China
- Prior art keywords
- domain
- target
- data
- preset
- fault category
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
Abstract
The invention discloses a multi-domain joint fault prediction method, a multi-domain joint fault prediction device, a multi-domain joint fault prediction terminal and a storage medium for heterogeneous equipment, wherein the method comprises the following steps: acquiring cross-domain historical data of target equipment before a target moment, and constructing target data sequences corresponding to all domains respectively; inputting each target data sequence into a trained first neural network, and extracting characteristic information corresponding to each domain in each preset fault category; fusing the feature information corresponding to each preset fault category of each domain to obtain fused feature information of each preset fault category, and acquiring a fault prediction result of the target equipment at the target moment according to the fused feature information; wherein the domain comprises a time domain, a frequency domain, a time-frequency domain and a graphic domain. The invention can realize accurate equipment fault prediction.
Description
Technical Field
The invention relates to the technical field of terminal control, in particular to a heterogeneous device multi-domain joint fault prediction method, a heterogeneous device multi-domain joint fault prediction device, a heterogeneous device multi-domain joint fault prediction terminal and a storage medium.
Background
The coverage range of the power system is very wide, and in order to maintain the normal operation of the power system, multiple devices are required to operate cooperatively, and the multiple devices are also required to acquire mass data for analysis. When equipment fails, the safety and stability of the whole system can be affected due to incomplete or unreliable process data, and no method for predicting the failure of the equipment in the system exists at present.
Thus, there is a need for improvements and enhancements in the art.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a method, a device, a terminal and a storage medium for predicting multi-domain joint faults of heterogeneous equipment, and aims to solve the problem that no method for predicting the faults of the equipment in a power system exists in the prior art.
In order to solve the technical problems, the technical scheme adopted by the invention is as follows:
in a first aspect of the present invention, a method for predicting a multi-domain joint failure of a heterogeneous device is provided, where the method includes:
acquiring cross-domain historical data of target equipment before a target moment, and constructing target data sequences corresponding to all domains respectively;
inputting each target data sequence into a trained first neural network, and extracting characteristic information corresponding to each domain in each preset fault category;
fusing the feature information corresponding to each preset fault category of each domain to obtain fused feature information of each preset fault category, and acquiring a fault prediction result of the target equipment at the target moment according to the fused feature information;
wherein the domain comprises a time domain, a frequency domain, a time-frequency domain and a graphic domain.
The method for predicting the multi-domain joint failure of the heterogeneous device, wherein the acquiring of cross-domain historical data of the target device before the target time and the constructing of target data sequences corresponding to each domain respectively, comprises:
acquiring historical data of each domain of the target equipment before the target moment, and processing abnormal data in the historical data;
and normalizing the historical data before the target time corresponding to each domain to construct the target data sequence corresponding to each domain.
The method for predicting the multi-domain joint failure of the heterogeneous device, wherein the processing of the abnormal data in the historical data comprises:
when the variation range of target data in the historical data relative to the adjacent data is larger than a second preset threshold value, determining that the target data is error data;
updating the error data into the average value of the front and back adjacent data;
for missing data in the historical data:
if the acquisition time interval of the front and back adjacent data of the missing data is not larger than a first preset threshold, taking the average value of the front and back adjacent data of the missing data as the supplementary data of the missing data;
and if the acquisition time interval of the adjacent data before and after the missing data is greater than the first preset threshold, searching the historical data for the data acquired at the time corresponding to the acquisition time of the missing data as the supplementary data of the missing data.
The method for predicting the multi-domain joint failure of the heterogeneous device, wherein the step of fusing the feature information corresponding to each preset failure category of each domain to obtain the fused feature information of each preset failure category comprises the following steps:
fusing the characteristic information corresponding to each preset fault category of each domain according to a preset formula;
the preset formula is as follows:
wherein the content of the first and second substances,
fusion characteristic information representing the jth of the preset fault categories,
representing historical data according to the time t before the target timeObtaining characteristic information corresponding to the jth domain in the jth preset fault category, wherein i is the number of the domains;
for the fusion parameter corresponding to the jth preset fault category at the target time t,
。
the multi-domain joint failure prediction method for the heterogeneous equipment,
predicting through a trained second neural network; before fusing the feature information corresponding to each preset fault category of each domain according to a preset formula, the method further includes:
acquiring a time interval between the target time and the starting operation time of the target equipment; inputting the characteristic information corresponding to the time interval and the jth preset fault category of each domain into the trained second neural network;
acquiring the fusion parameter corresponding to the jth preset fault category output by the second neural network at the target time t;
wherein the parameters of the second neural network are obtained by joint training with the first neural network.
The method for predicting the multi-domain joint failure of the heterogeneous device, wherein the obtaining of the failure prediction result of the target device at the target time according to the fusion characteristic information comprises:
inputting the fusion characteristic information of each preset fault category into a normalization index function;
and acquiring the probability corresponding to each preset fault category output by the normalization index function.
The method for predicting the multi-domain joint failure of the heterogeneous device, wherein before inputting each target data sequence into the trained first neural network, the method comprises:
training the first neural network and the second neural network according to multiple groups of sample data, wherein each group of sample data comprises sample data sequences respectively corresponding to each domain and constructed according to historical data of each domain of the target equipment before a preset moment, and a fault type corresponding to the target equipment at the preset moment;
the training the first neural network and the second neural network according to multiple sets of sample data comprises:
selecting target sample data from the multiple groups of sample data;
inputting the sample data sequence in target sample data into the first neural network, and extracting sample characteristic information corresponding to each preset fault category of each domain;
inputting the time interval between the preset time and the starting operation time of the target equipment and the sample characteristic information into the second neural network, and acquiring the fusion parameters, which are output by the second neural network and correspond to the preset time of each preset fault category in the target sample data;
fusing the sample characteristic information corresponding to each preset fault category in each domain at each preset fault category according to the fusion parameters corresponding to each preset fault category at the preset time in the target sample data to obtain sample fusion characteristic information of each preset fault category, and acquiring a fault prediction result of the target equipment at the preset time according to the sample fusion characteristic information;
acquiring a first loss according to the fault prediction result and the fault category in the target sample data;
acquiring a second loss according to the sample fusion characteristic information and the sample data sequence in the target sample data;
acquiring training loss corresponding to the target sample data according to the first loss and the second loss;
updating parameters of the first neural network and the second neural network according to the training loss;
re-executing the step of selecting target sample data in the plurality of sets of sample data until parameters of the first and second neural networks converge.
In a second aspect of the present invention, a device for predicting a multi-domain joint failure of a heterogeneous device is provided, including:
the target data sequence construction module is used for acquiring cross-domain historical data of the target equipment before a target moment and constructing target data sequences corresponding to all domains;
the characteristic information extraction module is used for inputting each target data sequence into a trained first neural network and extracting the characteristic information corresponding to each domain in each preset fault category;
the characteristic fusion and prediction module is used for fusing characteristic information corresponding to each preset fault category of each domain to obtain fusion characteristic information of each preset fault category, and acquiring a fault prediction result of the target equipment at the target moment according to the fusion characteristic information;
wherein the domain comprises a time domain, a frequency domain, a time-frequency domain and a graphic domain.
In a third aspect of the present invention, a terminal is provided, where the terminal includes a processor, and a computer-readable storage medium communicatively connected to the processor, where the computer-readable storage medium is adapted to store a plurality of instructions, and the processor is adapted to call the instructions in the computer-readable storage medium to perform steps implementing the multi-domain joint failure prediction method for a heterogeneous device according to any one of the foregoing descriptions.
In a fourth aspect of the present invention, a computer-readable storage medium is provided, which stores one or more programs, which are executable by one or more processors, to implement the steps of the heterogeneous device multi-domain joint failure prediction method described in any one of the above.
Compared with the prior art, the invention provides a multi-domain joint failure prediction method, a device, a terminal and a storage medium for heterogeneous equipment, constructing a target data sequence corresponding to each domain according to historical data of target equipment in a plurality of months before a target moment, inputting each target data sequence into a trained first neural network, extracting characteristic information corresponding to each domain in each fault category, fusing the characteristic information corresponding to each domain in each preset fault category to obtain fused characteristic information of each preset fault category, according to the method, the fault prediction result of the target equipment at the target moment is obtained according to the fusion characteristic information, the multi-domain characteristics are extracted and fused through the trained first neural network model to be used for fault prediction, and accurate equipment fault prediction can be achieved.
Drawings
Fig. 1 is a flowchart of an embodiment of a multi-domain joint failure prediction method for a heterogeneous device according to the present invention;
FIG. 2 is a schematic diagram of an LSTM network;
fig. 3 is a schematic structural diagram of an embodiment of a multi-domain joint failure prediction apparatus for heterogeneous devices according to the present invention;
fig. 4 is a schematic diagram of an embodiment of a terminal according to the present invention.
Detailed Description
In order to make the objects, technical solutions and effects of the present invention clearer and clearer, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The multi-domain joint fault prediction method for the heterogeneous equipment can be applied to a terminal with computing capacity, the terminal can execute the multi-domain joint fault prediction method for the heterogeneous equipment to perform fault prediction on target equipment, and the terminal can be but is not limited to various computers, mobile terminals, intelligent household appliances, wearable equipment and the like.
Example one
As shown in fig. 1, in an embodiment of the method for predicting the multi-domain joint failure of the heterogeneous device, the method includes the steps of:
s100, cross-domain historical data of the target equipment before the target moment is obtained, and target data sequences corresponding to all domains are constructed.
In order to predict the possibility of a failure of the target device at a target time, the target device may be a sensor, and the like, first obtaining cross-domain historical data of the target device before the target time, and constructing a target data sequence corresponding to each domain, specifically, each domain may include a time domain, a frequency domain, a time-frequency domain, and a graph domain, and obtaining cross-domain historical data of the target device before the target time, and constructing a target data sequence corresponding to each domain includes:
acquiring historical data of each domain of the target equipment before the target moment, and processing abnormal data in the historical data;
and normalizing the historical data before the target time corresponding to each domain to construct the target data sequence corresponding to each domain.
The historical data of each domain of the target device before the target time may be obtained through a working log of the target device, or obtained through a collecting device, such as a sensor, connected with or disposed inside the target device. The historical data of each domain of the target device before the target time is data between the target time and a preset time before the target time, and may be operation data, deformation data, temperature data, and the like. The target device may have abnormal data, such as data missing, error data, and the like, in the historical data acquired before the target time, and the target data sequence corresponding to each domain needs to be reconstructed after the abnormal data in the historical data is processed.
The processing of the abnormal data in the historical data comprises:
for missing data in the historical data:
if the acquisition time interval of the front and back adjacent data of the missing data is not larger than a first preset threshold, taking the average value of the front and back adjacent data of the missing data as the supplementary data of the missing data;
and if the acquisition time interval of the adjacent data before and after the missing data is greater than the first preset threshold, searching the historical data for the data acquired at the time corresponding to the acquisition time of the missing data as the supplementary data of the missing data.
Generally, the device automatically acquires data at certain intervals, so that whether missing data exists or not can be determined according to the interval of the acquisition time of other data, or whether missing data exists or not is determined when the data acquisition time exists but no corresponding specific data exists, whether the acquisition time interval of the two data is greater than a first preset threshold value is determined when the missing data exists between the two data, if the missing data does not exist, the average value of the two data is used as supplementary data of the missing data, and if the missing data exists, the data which is the same as the missing data and corresponds to the acquisition time is searched in other historical data and is used as supplementary data of the missing data, and the time correspondence refers to the same position in the same time period.
In a possible implementation manner, the processing abnormal data in the historical data includes:
when the variation range of target data in the historical data relative to the adjacent data is larger than a second preset threshold value, determining that the target data is error data;
and updating the error data into the average value of the front and back adjacent data.
Between the acquisition time of the target data and the acquisition time of the adjacent data, when the variation range of the target data in the historical data relative to the adjacent data is too large, the target data is possibly interfered during acquisition, so that data errors are caused, and the error data is updated to be the average value of the adjacent data, so that the influence of the error data on the prediction result is reduced.
And after processing the historical data of the target equipment in each domain before the target time, normalizing the processed historical data of each domain before the target time to construct each target data sequence corresponding to the processed historical data.
The processed historical data of the target device in each domain is a data sequence, and comprises data collected at each moment before the target moment, and a formula is adopted:
carrying out normalization processing to obtain a data sequence after normalization processing as the target data sequence, namely, each domain corresponds to one target data sequence, wherein,
representing the m-th data in the target data sequence corresponding to the i-th domain,
representing the mth data in the processed history data sequence of the ith field,
represents the largest data in the processed historical data sequence of the ith domain,
indicating the minimum data in the processed historical data sequence of the ith domain.
Referring to fig. 1 again, the method for predicting multi-domain joint failure of heterogeneous devices according to the present embodiment further includes the following steps:
s200, inputting each target data sequence into the trained first neural network, and extracting the characteristic information corresponding to each domain in each preset fault category.
Specifically, the first neural network may be constructed based on an existing network structure, for example, the first neural network may be an LSTM network, a specific structure of the LSTM network is shown in fig. 2, and of course, other structures of the first neural network, for example, a structure of a CNN and a residual error network, may also be adopted to construct the first neural network. After the first neural network is trained, the function of outputting the characteristic information corresponding to each preset fault category according to a plurality of input data sequences can be realized, and the characteristic information corresponding to each preset fault category output by the first neural network can be used for acquiring a corresponding fault prediction result after fusion processing.
The training process of the first neural network will be explained later. After obtaining target data sequences corresponding to the target devices in the domains respectively, inputting the target data sequence corresponding to each domain into the trained first neural network, and obtaining feature information corresponding to each preset fault type of each domain output by the first neural network, that is, for each domain, a feature information group is generated, where the feature information group includes feature information corresponding to the domain in different preset fault types.
The preset fault category may be determined based on historical fault data.
S300, fusing the feature information corresponding to each preset fault category of each domain to obtain the fused feature information of each preset fault category, and acquiring the fault prediction result of the target equipment at the target moment according to the fused feature information.
After obtaining the feature information of each domain corresponding to each preset fault category, fusing the feature information of each domain corresponding to each preset fault category, in this embodiment, specifically, fusing the feature information of each domain corresponding to each preset fault category to obtain the fused feature information of each preset fault category includes:
fusing the characteristic information corresponding to each preset fault category of each domain according to a preset formula;
the preset formula is as follows:
wherein the content of the first and second substances,
fusion characteristic information representing the jth of the preset fault categories,
representing characteristic information corresponding to the jth preset fault category of the ith domain obtained according to historical data before the target time t, wherein i is the number of the domains;
for the fusion parameter corresponding to the jth preset fault category at the target time t,
。
acquiring a time interval between the target time and the starting operation time of the target equipment; inputting the characteristic information corresponding to the time interval and the jth preset fault category of each domain into the trained second neural network;
acquiring the fusion parameter corresponding to the jth preset fault category output by the second neural network at the target time t;
wherein the parameters of the second neural network are obtained by joint training with the first neural network.
The joint training process of the first neural network and the second neural network will be described later.
Fusing the feature information of each domain in the same preset fault category to obtain fused feature information corresponding to each preset fault category, wherein obtaining the fault prediction result of the target device at the target time according to the fused feature information comprises:
inputting the fusion characteristic information of each preset fault category into a normalization index function;
and acquiring the probability corresponding to each preset fault category output by the normalization index function.
Specifically, the normalization index function is also called a softmax function, and is used for performing normalization calculation on a plurality of input data. Inputting the fusion characteristic information of each preset fault category into a softmax function to obtain the probability corresponding to each preset fault category, namely the probability that each preset fault category exists in the target device at the target moment, and taking the probability as the fault prediction result of the target device at the target moment.
The joint training process of the first and second neural networks is described below:
before inputting each target data sequence into the trained first neural network, the method includes:
training the first neural network and the second neural network according to multiple groups of sample data, wherein each group of sample data comprises sample data sequences corresponding to each domain and constructed according to historical data of each domain of the target equipment before a preset time, and fault types corresponding to the target equipment at the preset time.
Firstly, a plurality of groups of sample data are constructed for training the first neural network, in order to ensure the accuracy of the first neural network when the feature information output according to the input data sequence is used for fault type prediction, each group of sample data comprises a sample data sequence, the sample data sequence is a sample data sequence corresponding to each domain constructed according to the historical data of each domain of the target device before the preset time, that is, the process of constructing the sample data sequence can be performed with reference to the step S100, that is, after abnormal data in the cross-domain historical data of the target device before the preset time is processed, the historical data of each domain is normalized to obtain the sample data sequence corresponding to each domain. Each sample data also comprises a fault category corresponding to the preset time adopted by the target equipment for constructing the sample data sequence in the sample data, and the preset time in different sample data is different. So that the failure occurrence mechanism of the target device in different operation stages can be learned.
The training the first neural network and the second neural network according to multiple sets of sample data comprises:
selecting target sample data from the multiple groups of sample data;
inputting the sample data sequence in target sample data into the first neural network, and extracting sample characteristic information corresponding to each preset fault category of each domain;
inputting the time interval between the preset time and the starting operation time of the target equipment and the sample characteristic information into the second neural network, and acquiring the fusion parameters, which are output by the second neural network and correspond to the preset time of each preset fault category in the target sample data;
fusing the sample characteristic information corresponding to each preset fault category in each domain at each preset fault category according to the fusion parameters corresponding to each preset fault category at the preset time in the target sample data to obtain sample fusion characteristic information of each preset fault category, and acquiring a fault prediction result of the target equipment at the preset time according to the sample fusion characteristic information;
acquiring a first loss according to the fault prediction result and the fault category in the target sample data;
acquiring a second loss according to the sample fusion characteristic information and the sample data sequence in the target sample data;
acquiring training loss corresponding to the target sample data according to the first loss and the second loss;
updating parameters of the first neural network and the second neural network according to the training loss;
re-executing the step of selecting target sample data in the plurality of sets of sample data until parameters of the first and second neural networks converge.
When the first neural network and the second neural network are trained, firstly, one group of sample data is selected from multiple groups of sample data to be used as target sample data, the sample data sequence in the target sample data is input into the first neural network, and sample characteristic information corresponding to each domain in each preset fault category is extracted through the first neural network; then, inputting the time interval between the preset time in the target sample data and the start operation time of the target device and the sample characteristic information into the second neural network, obtaining the fusion parameters output by the second neural network, that is, obtaining the fusion parameters corresponding to the preset time of each preset fault category in the target sample data, then fusing the sample characteristic information corresponding to each preset fault category in each domain according to the corresponding fusion parameters, where the fusion formula may refer to the formula in step S300, so as to obtain the sample fusion characteristic information of each preset fault category, then obtaining the fault prediction result of the target device at the preset time according to the sample fusion characteristic information, and calculating a first loss according to the difference between the fault prediction result and the fault category in the target sample data, it is easy to see that, since the fault category in the target sample data is real fault information, the smaller the first loss is, the better the first loss is, the parameters of the first neural network and the second neural network can be updated under the constraint that the first loss is minimal, so that the first neural network and the second neural network can realize accurate fault prediction.
In order to improve the efficiency of the first neural network and the second neural network, when the feature information corresponding to each domain in each preset fault category is extracted through the first neural network and the feature information is fused through the second neural network, the feature information in the initial data can be reserved and utilized as much as possible, thus, further, in addition to calculating the first loss, a difference between the sample fusion feature information and the sample data sequence in the target sample data is calculated as a second loss, the smaller the second loss, the better, the characteristic information in the sample data sequence can be reserved and utilized as much as possible, in this embodiment, the parameters of the first and second neural networks are updated under the constraint that the sum of the first and second losses is minimal.
And after the parameters of the first neural network and the second neural network are updated once by adopting one target sample data each time, reselecting one sample data as new target sample data, then updating the parameters of the first neural network and the second neural network, after multiple rounds of updating, converging the parameters of the first neural network and the second neural network, finishing training, and using the first neural network and the second neural network with converged parameters for the fault prediction of the target equipment.
In summary, this embodiment provides a multi-domain joint fault prediction method for heterogeneous devices, where a target data sequence corresponding to each domain is constructed according to historical data of a target device in multiple months before a target time, and each target data sequence is input to a trained first neural network, feature information corresponding to each fault category of each domain is extracted, and feature information corresponding to each preset fault category of each domain is fused to obtain fused feature information of each preset fault category, and a fault prediction result of the target device at the target time is obtained according to the fused feature information.
It should be understood that, although the steps in the flowcharts shown in the figures of the present specification are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least a portion of the steps in the flowchart may include multiple sub-steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of performing the sub-steps or stages is not necessarily sequential, but may be performed alternately or alternately with other steps or at least a portion of the sub-steps or stages of other steps.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, databases, or other media used in embodiments provided herein may include non-volatile and/or volatile memory. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
Example two
Based on the foregoing embodiment, the present invention further provides a multi-domain joint failure prediction apparatus for heterogeneous devices, where as shown in fig. 3, the multi-domain joint failure prediction apparatus for heterogeneous devices includes:
a target data sequence construction module, configured to obtain cross-domain historical data of a target device before a target time, and construct a target data sequence corresponding to each domain, as described in embodiment one;
a feature information extraction module, configured to input each target data sequence to a trained first neural network, and extract feature information corresponding to each domain in each preset fault category, as described in embodiment one;
the feature fusion and prediction module is configured to fuse feature information corresponding to each preset fault category of each domain to obtain fusion feature information of each preset fault category, and obtain a fault prediction result of the target device at the target time according to the fusion feature information, which is specifically described in embodiment one;
the domain includes a time domain, a frequency domain, a time-frequency domain, and a graphic domain, which is specifically described in the first embodiment.
EXAMPLE III
Based on the above embodiments, the present invention further provides a terminal, as shown in fig. 4, where the terminal includes a processor 10 and a memory 20. Fig. 4 shows only some of the components of the terminal, but it is to be understood that not all of the shown components are required to be implemented, and that more or fewer components may be implemented instead.
The memory 20 may in some embodiments be an internal storage unit of the terminal, such as a hard disk or a memory of the terminal. The memory 20 may also be an external storage device of the terminal in other embodiments, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), and the like provided on the terminal. Further, the memory 20 may also include both an internal storage unit and an external storage device of the terminal. The memory 20 is used for storing application software installed in the terminal and various data. The memory 20 may also be used to temporarily store data that has been output or is to be output. In an embodiment, the memory 20 stores a multi-domain joint failure prediction program 30 for a heterogeneous device, and the multi-domain joint failure prediction program 30 is executable by the processor 10, so as to implement the multi-domain joint failure prediction method for a heterogeneous device in the present application.
The processor 10 may be a Central Processing Unit (CPU), a microprocessor or other chip in some embodiments, and is configured to run program codes stored in the memory 20 or process data, for example, execute the heterogeneous device multi-domain joint failure prediction method.
In one embodiment, when processor 10 executes heterogeneous device multi-domain joint failure prediction program 30 in memory 20, the following steps are implemented:
acquiring cross-domain historical data of target equipment before a target moment, and constructing target data sequences corresponding to all domains respectively;
inputting each target data sequence into a trained first neural network, and extracting characteristic information corresponding to each domain in each preset fault category;
fusing the feature information corresponding to each preset fault category of each domain to obtain fused feature information of each preset fault category, and acquiring a fault prediction result of the target equipment at the target moment according to the fused feature information;
wherein the domain comprises a time domain, a frequency domain, a time-frequency domain and a graphic domain.
The acquiring cross-domain historical data of the target device before the target moment and constructing target data sequences corresponding to each domain respectively comprises:
acquiring historical data of each domain of the target equipment before the target moment, and processing abnormal data in the historical data;
and normalizing the historical data before the target time corresponding to each domain to construct the target data sequence corresponding to each domain.
Wherein, the processing the abnormal data in the historical data comprises:
when the variation range of target data in the historical data relative to the adjacent data is larger than a second preset threshold value, determining that the target data is error data;
updating the error data into the average value of the front and back adjacent data;
for missing data in the historical data:
if the acquisition time interval of the front and back adjacent data of the missing data is not larger than a first preset threshold, taking the average value of the front and back adjacent data of the missing data as the supplementary data of the missing data;
and if the acquisition time interval of the adjacent data before and after the missing data is greater than the first preset threshold, searching the historical data for the data acquired at the time corresponding to the acquisition time of the missing data as the supplementary data of the missing data.
The fusing the feature information corresponding to each preset fault category of each domain to obtain the fused feature information of each preset fault category includes:
fusing the characteristic information corresponding to each preset fault category of each domain according to a preset formula;
the preset formula is as follows:
wherein the content of the first and second substances,
fusion characteristic information representing the jth of the preset fault categories,
representing characteristic information corresponding to the jth preset fault category of the ith domain obtained according to historical data before the target time t, wherein i is the number of the domains;
for the fusion parameter corresponding to the jth preset fault category at the target time t,
。
wherein the content of the first and second substances,
predicting through a trained second neural network; before fusing the feature information corresponding to each preset fault category of each domain according to a preset formula, the method further includes:
acquiring a time interval between the target time and the starting operation time of the target equipment; inputting the characteristic information corresponding to the time interval and the jth preset fault category of each domain into the trained second neural network;
acquiring the fusion parameter corresponding to the jth preset fault category output by the second neural network at the target time t;
wherein the parameters of the second neural network are obtained by joint training with the first neural network.
The obtaining of the failure prediction result of the target device at the target time according to the fusion feature information includes:
inputting the fusion characteristic information of each preset fault category into a normalization index function;
and acquiring the probability corresponding to each preset fault category output by the normalization index function.
Wherein before inputting each of the target data sequences into the trained first neural network, the method comprises:
training the first neural network and the second neural network according to multiple groups of sample data, wherein each group of sample data comprises sample data sequences respectively corresponding to each domain and constructed according to historical data of each domain of the target equipment before a preset moment, and a fault type corresponding to the target equipment at the preset moment;
the training the first neural network and the second neural network according to multiple sets of sample data comprises:
selecting target sample data from the multiple groups of sample data;
inputting the sample data sequence in target sample data into the first neural network, and extracting sample characteristic information corresponding to each preset fault category of each domain;
inputting the time interval between the preset time and the starting operation time of the target equipment and the sample characteristic information into the second neural network, and acquiring the fusion parameters, which are output by the second neural network and correspond to the preset time of each preset fault category in the target sample data;
fusing the sample characteristic information corresponding to each preset fault category in each domain at each preset fault category according to the fusion parameters corresponding to each preset fault category at the preset time in the target sample data to obtain sample fusion characteristic information of each preset fault category, and acquiring a fault prediction result of the target equipment at the preset time according to the sample fusion characteristic information;
acquiring a first loss according to the fault prediction result and the fault category in the target sample data;
acquiring a second loss according to the sample fusion characteristic information and the sample data sequence in the target sample data;
acquiring training loss corresponding to the target sample data according to the first loss and the second loss;
updating parameters of the first neural network and the second neural network according to the training loss;
re-executing the step of selecting target sample data in the plurality of sets of sample data until parameters of the first and second neural networks converge.
Example four
The present invention also provides a computer readable storage medium, in which one or more programs are stored, the one or more programs being executable by one or more processors to implement the steps of the heterogeneous device multi-domain joint failure prediction method as described above.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (9)
1. A multi-domain joint failure prediction method for heterogeneous equipment is characterized by comprising the following steps:
acquiring cross-domain historical data of target equipment before a target moment, and constructing target data sequences corresponding to all domains respectively;
inputting each target data sequence into a trained first neural network, and extracting characteristic information corresponding to each domain in each preset fault category;
fusing the feature information corresponding to each preset fault category of each domain to obtain fused feature information of each preset fault category, and acquiring a fault prediction result of the target equipment at the target moment according to the fused feature information;
wherein the domain comprises a time domain, a frequency domain, a time-frequency domain and a graphic domain;
the fusing the feature information corresponding to each preset fault category of each domain to obtain the fused feature information of each preset fault category comprises:
fusing the characteristic information corresponding to each preset fault category of each domain according to a preset formula;
the preset formula is as follows:
wherein the content of the first and second substances,
fusion characteristic information representing the jth of the preset fault categories,
representing characteristic information corresponding to the jth preset fault category of the ith domain obtained according to historical data before the target time t, wherein i is the number of the domains;
for the fusion parameter corresponding to the jth preset fault category at the target time t,
。
2. the method for predicting the multi-domain joint failure of the heterogeneous device according to claim 1, wherein the obtaining of cross-domain historical data of the target device before the target time and the construction of target data sequences corresponding to each domain respectively comprises:
acquiring historical data of each domain of the target equipment before the target moment, and processing abnormal data in the historical data;
and normalizing the historical data before the target time corresponding to each domain to construct the target data sequence corresponding to each domain.
3. The multi-domain joint failure prediction method for the heterogeneous device according to claim 2, wherein the processing of the abnormal data in the historical data comprises:
when the variation range of target data in the historical data relative to the adjacent data is larger than a second preset threshold value, determining that the target data is error data;
updating the error data into the average value of the front and back adjacent data;
for missing data in the historical data:
if the acquisition time interval of the front and back adjacent data of the missing data is not larger than a first preset threshold, taking the average value of the front and back adjacent data of the missing data as the supplementary data of the missing data;
and if the acquisition time interval of the adjacent data before and after the missing data is greater than the first preset threshold, searching the historical data for the data acquired at the time corresponding to the acquisition time of the missing data as the supplementary data of the missing data.
4. The multi-domain joint failure prediction method of heterogeneous devices according to claim 1,
predicting through a trained second neural network; before fusing the feature information corresponding to each preset fault category of each domain according to a preset formula, the method further includes:
acquiring a time interval between the target time and the starting operation time of the target equipment; inputting the characteristic information corresponding to the time interval and the jth preset fault category of each domain into the trained second neural network;
acquiring the fusion parameter corresponding to the jth preset fault category output by the second neural network at the target time t;
wherein the parameters of the second neural network are obtained by joint training with the first neural network.
5. The method for predicting the multi-domain joint failure of the heterogeneous device according to claim 4, wherein the obtaining of the failure prediction result of the target device at the target time according to the fusion feature information comprises:
inputting the fusion characteristic information of each preset fault category into a normalization index function;
and acquiring the probability corresponding to each preset fault category output by the normalization index function.
6. The method according to claim 5, wherein before inputting each target data sequence into the trained first neural network, the method comprises:
training the first neural network and the second neural network according to multiple groups of sample data, wherein each group of sample data comprises sample data sequences respectively corresponding to each domain and constructed according to historical data of each domain of the target equipment before a preset moment, and a fault type corresponding to the target equipment at the preset moment;
the training the first neural network and the second neural network according to multiple sets of sample data comprises:
selecting target sample data from the multiple groups of sample data;
inputting the sample data sequence in target sample data into the first neural network, and extracting sample characteristic information corresponding to each preset fault category of each domain;
inputting the time interval between the preset time and the starting operation time of the target equipment and the sample characteristic information into the second neural network, and acquiring the fusion parameters, which are output by the second neural network and correspond to the preset time of each preset fault category in the target sample data;
fusing the sample characteristic information corresponding to each preset fault category in each domain at each preset fault category according to the fusion parameters corresponding to each preset fault category at the preset time in the target sample data to obtain sample fusion characteristic information of each preset fault category, and acquiring a fault prediction result of the target equipment at the preset time according to the sample fusion characteristic information;
acquiring a first loss according to the fault prediction result and the fault category in the target sample data;
acquiring a second loss according to the sample fusion characteristic information and the sample data sequence in the target sample data;
acquiring training loss corresponding to the target sample data according to the first loss and the second loss;
updating parameters of the first neural network and the second neural network according to the training loss;
re-executing the step of selecting target sample data in the plurality of sets of sample data until parameters of the first and second neural networks converge.
7. A multi-domain joint failure prediction device for heterogeneous equipment is characterized by comprising:
the target data sequence construction module is used for acquiring cross-domain historical data of the target equipment before a target moment and constructing target data sequences corresponding to all domains;
the characteristic information extraction module is used for inputting each target data sequence into a trained first neural network and extracting the characteristic information corresponding to each domain in each preset fault category;
the characteristic fusion and prediction module is used for fusing characteristic information corresponding to each preset fault category of each domain to obtain fusion characteristic information of each preset fault category, and acquiring a fault prediction result of the target equipment at the target moment according to the fusion characteristic information;
wherein the domain comprises a time domain, a frequency domain, a time-frequency domain and a graphic domain;
the fusing the feature information corresponding to each preset fault category of each domain to obtain the fused feature information of each preset fault category comprises:
fusing the characteristic information corresponding to each preset fault category of each domain according to a preset formula;
the preset formula is as follows:
wherein the content of the first and second substances,
fusion characteristic information representing the jth of the preset fault categories,
representing characteristic information corresponding to the jth preset fault category of the ith domain obtained according to historical data before the target time t, wherein i is the number of the domains;
for the fusion parameter corresponding to the jth preset fault category at the target time t,
。
8. a terminal, characterized in that the terminal comprises: a processor, a computer readable storage medium communicatively connected to the processor, the computer readable storage medium adapted to store a plurality of instructions, the processor adapted to invoke the instructions in the computer readable storage medium to perform the steps of implementing the heterogeneous device multi-domain joint failure prediction method of any of the above claims 1-6.
9. A computer-readable storage medium, storing one or more programs, the one or more programs being executable by one or more processors to perform the steps of the heterogeneous device multi-domain joint failure prediction method according to any one of claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111558674.XA CN113935443B (en) | 2021-12-20 | 2021-12-20 | Heterogeneous equipment multi-domain joint fault prediction method, device, terminal and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111558674.XA CN113935443B (en) | 2021-12-20 | 2021-12-20 | Heterogeneous equipment multi-domain joint fault prediction method, device, terminal and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113935443A CN113935443A (en) | 2022-01-14 |
| CN113935443B true CN113935443B (en) | 2022-04-22 |
Family
ID=79289239
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111558674.XA Active CN113935443B (en) | 2021-12-20 | 2021-12-20 | Heterogeneous equipment multi-domain joint fault prediction method, device, terminal and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113935443B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114675597B (en) * | 2022-05-30 | 2022-08-26 | 中科航迈数控软件(深圳)有限公司 | Fault prediction visualization method for numerical control machine tool |
| CN115293057B (en) * | 2022-10-10 | 2022-12-20 | 深圳先进技术研究院 | Wind driven generator fault prediction method based on multi-source heterogeneous data |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112291807A (en) * | 2020-10-15 | 2021-01-29 | 山东科技大学 | Wireless cellular network traffic prediction method based on deep migration learning and cross-domain data fusion |
| CN113342476A (en) * | 2021-07-05 | 2021-09-03 | 中山大学 | Equipment cluster cross-domain fault prediction method and system based on transfer learning |
| CN113592019A (en) * | 2021-08-10 | 2021-11-02 | 平安银行股份有限公司 | Fault detection method, device, equipment and medium based on multi-model fusion |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3480714A1 (en) * | 2017-11-03 | 2019-05-08 | Tata Consultancy Services Limited | Signal analysis systems and methods for features extraction and interpretation thereof |
| CN109186973B (en) * | 2018-08-20 | 2020-03-31 | 东南大学 | Mechanical fault diagnosis method of unsupervised deep learning network |
| US11416357B2 (en) * | 2020-03-06 | 2022-08-16 | Dell Products L.P. | Method and system for managing a spare fault domain in a multi-fault domain data cluster |
-
2021
- 2021-12-20 CN CN202111558674.XA patent/CN113935443B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112291807A (en) * | 2020-10-15 | 2021-01-29 | 山东科技大学 | Wireless cellular network traffic prediction method based on deep migration learning and cross-domain data fusion |
| CN113342476A (en) * | 2021-07-05 | 2021-09-03 | 中山大学 | Equipment cluster cross-domain fault prediction method and system based on transfer learning |
| CN113592019A (en) * | 2021-08-10 | 2021-11-02 | 平安银行股份有限公司 | Fault detection method, device, equipment and medium based on multi-model fusion |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113935443A (en) | 2022-01-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111210024B (en) | Model training method, device, computer equipment and storage medium | |
| CN113935443B (en) | Heterogeneous equipment multi-domain joint fault prediction method, device, terminal and storage medium | |
| AU2019201857B2 (en) | Sparse neural network based anomaly detection in multi-dimensional time series | |
| CN111832294B (en) | Method and device for selecting marking data, computer equipment and storage medium | |
| CN111652615B (en) | Safety identification method based on block chain big data and artificial intelligence cloud service platform | |
| CN111625516A (en) | Method and device for detecting data state, computer equipment and storage medium | |
| CN115174231B (en) | Network fraud analysis method and server based on AI Knowledge Base | |
| CN110069364B (en) | Spare part data error correction method and device, computer equipment and storage medium | |
| CN115656673A (en) | Transformer data processing device and equipment storage medium | |
| CN115293057B (en) | Wind driven generator fault prediction method based on multi-source heterogeneous data | |
| CN109271957B (en) | Face gender identification method and device | |
| US20220215293A1 (en) | Method to identify incorrect account numbers | |
| CN115800272A (en) | Power grid fault analysis method, system, terminal and medium based on topology identification | |
| CN111124898B (en) | Question-answering system testing method and device, computer equipment and storage medium | |
| CN110391840B (en) | Method and system for judging abnormality of telemetry parameters of sun synchronous orbit satellite | |
| CN113110961B (en) | Equipment abnormality detection method and device, computer equipment and readable storage medium | |
| CN111885181B (en) | Monitoring data reporting method and device, computer equipment and storage medium | |
| CN117057794A (en) | Block chain-based intelligent payment gateway transaction method, system, equipment and medium | |
| CN116450137A (en) | System abnormality detection method and device, storage medium and electronic equipment | |
| CN114386980A (en) | Fraud identification method and device based on time sequence data driving | |
| CN110865939B (en) | Application program quality monitoring method, device, computer equipment and storage medium | |
| CN112215067A (en) | Protection action analysis method and device, electronic equipment and storage medium | |
| CN115373868A (en) | Intelligent data processing system and method | |
| CN112883590A (en) | System reliability verification test scheme determination method | |
| CN115965324B (en) | Commodity sales method and system based on vending machine |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |